[rrd-users] Re: rrdtool resets (sorry)
don.baarda at baesystems.com
Mon Jan 29 00:19:40 MET 2001
> -----Original Message-----
> From: Christian Pearce [SMTP:pearcec at PacketPushers.com]
> Sent: Saturday, January 27, 2001 2:57 PM
> To: Paul Wickman
> Cc: Christian Pearce; rrd-users at list.ee.ethz.ch
> Subject: [rrd-users] Re: rrdtool resets (sorry)
> Lookin at the code and following the values during a reset. How does it
> calculate the interval that had the reset occur during it. I thought the
> interval would go NAN since the delta should be negative. But it looks
> like it calculates a value.
Remember that counters can "zero" because of a wrap, not just a
reset, and RRD has no way to differentiate the two. So RRD treats a negative
delta for a COUNTER as a wrap by default. If it was actually a reset, this
would _normaly_ result in an unlikely value (ie, huge). So the only way RRD
can mark a reset interval as "UNKNOWN" is if the value for that interval
looks unreasonable; ie, too big.
> I take it the DERIVE is the right thing to do and I believe in my mind it
> is correct. I just would like to be able to explain why it is correct to
> my boss and everyone else that asks. Any takers?
I haven't seen all of your earlier posts so I'm not 100% familiar
with your setup, but it sounds like what you actually want is a COUNTER with
a reasonable max. Using DERIVE with min=0.0 will mark not only resets, but
also legitimate counter wraps as "Unknown".
It always pays to have reasonable min and max values.
> Christian Pearce
> On Fri, 26 Jan 2001, Paul Wickman wrote:
> > Use DERIVE instead of COUNTER. That, in combination with rrd-min=0 will
> > eliminate the "reset effect". You will loose the offending sample, but
> > my (and many other people's) case, it's a worthty tradeoff.
> Unsubscribe mailto:rrd-users-request at list.ee.ethz.ch?subject=unsubscribe
> Help mailto:rrd-users-request at list.ee.ethz.ch?subject=help
> Archive http://www.ee.ethz.ch/~slist/rrd-users
> WebAdmin http://www.ee.ethz.ch/~slist/lsg2.cgi
Unsubscribe mailto:rrd-users-request at list.ee.ethz.ch?subject=unsubscribe
Help mailto:rrd-users-request at list.ee.ethz.ch?subject=help
More information about the rrd-users